sdk/python/arvados/commands/put.py

   1 # Copyright (C) The Arvados Authors. All rights reserved.
   2 #
   3 # SPDX-License-Identifier: Apache-2.0
   4
   5 from __future__ import division
   6 from future.utils import listitems, listvalues
   7 from builtins import str
   8 from builtins import object
   9 import argparse
  10 import arvados
  11 import arvados.collection
  12 import base64
  13 import copy
  14 import datetime
  15 import errno
  16 import fcntl
  17 import fnmatch
  18 import hashlib
  19 import json
  20 import logging
  21 import os
  22 import pwd
  23 import re
  24 import signal
  25 import socket
  26 import sys
  27 import tempfile
  28 import threading
  29 import time
  30 import traceback
  31
  32 from apiclient import errors as apiclient_errors
  33 from arvados._version import __version__
  34 from arvados.util import keep_locator_pattern
  35
  36 import arvados.commands._util as arv_cmd
  37
  38 api_client = None
  39
  40 upload_opts = argparse.ArgumentParser(add_help=False)
  41
  42 upload_opts.add_argument('--version', action='version',
  43                          version="%s %s" % (sys.argv[0], __version__),
  44                          help='Print version and exit.')
  45 upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
  46                          help="""
  47 Local file or directory. If path is a directory reference with a trailing
  48 slash, then just upload the directory's contents; otherwise upload the
  49 directory itself. Default: read from standard input.
  50 """)
  51
  52 _group = upload_opts.add_mutually_exclusive_group()
  53
  54 _group.add_argument('--max-manifest-depth', type=int, metavar='N',
  55                     default=-1, help=argparse.SUPPRESS)
  56
  57 _group.add_argument('--normalize', action='store_true',
  58                     help="""
  59 Normalize the manifest by re-ordering files and streams after writing
  60 data.
  61 """)
  62
  63 _group.add_argument('--dry-run', action='store_true', default=False,
  64                     help="""
  65 Don't actually upload files, but only check if any file should be
  66 uploaded. Exit with code=2 when files are pending for upload.
  67 """)
  68
  69 _group = upload_opts.add_mutually_exclusive_group()
  70
  71 _group.add_argument('--as-stream', action='store_true', dest='stream',
  72                     help="""
  73 Synonym for --stream.
  74 """)
  75
  76 _group.add_argument('--stream', action='store_true',
  77                     help="""
  78 Store the file content and display the resulting manifest on
  79 stdout. Do not write the manifest to Keep or save a Collection object
  80 in Arvados.
  81 """)
  82
  83 _group.add_argument('--as-manifest', action='store_true', dest='manifest',
  84                     help="""
  85 Synonym for --manifest.
  86 """)
  87
  88 _group.add_argument('--in-manifest', action='store_true', dest='manifest',
  89                     help="""
  90 Synonym for --manifest.
  91 """)
  92
  93 _group.add_argument('--manifest', action='store_true',
  94                     help="""
  95 Store the file data and resulting manifest in Keep, save a Collection
  96 object in Arvados, and display the manifest locator (Collection uuid)
  97 on stdout. This is the default behavior.
  98 """)
  99
 100 _group.add_argument('--as-raw', action='store_true', dest='raw',
 101                     help="""
 102 Synonym for --raw.
 103 """)
 104
 105 _group.add_argument('--raw', action='store_true',
 106                     help="""
 107 Store the file content and display the data block locators on stdout,
 108 separated by commas, with a trailing newline. Do not store a
 109 manifest.
 110 """)
 111
 112 upload_opts.add_argument('--update-collection', type=str, default=None,
 113                          dest='update_collection', metavar="UUID", help="""
 114 Update an existing collection identified by the given Arvados collection
 115 UUID. All new local files will be uploaded.
 116 """)
 117
 118 upload_opts.add_argument('--use-filename', type=str, default=None,
 119                          dest='filename', help="""
 120 Synonym for --filename.
 121 """)
 122
 123 upload_opts.add_argument('--filename', type=str, default=None,
 124                          help="""
 125 Use the given filename in the manifest, instead of the name of the
 126 local file. This is useful when "-" or "/dev/stdin" is given as an
 127 input file. It can be used only if there is exactly one path given and
 128 it is not a directory. Implies --manifest.
 129 """)
 130
 131 upload_opts.add_argument('--portable-data-hash', action='store_true',
 132                          help="""
 133 Print the portable data hash instead of the Arvados UUID for the collection
 134 created by the upload.
 135 """)
 136
 137 upload_opts.add_argument('--replication', type=int, metavar='N', default=None,
 138                          help="""
 139 Set the replication level for the new collection: how many different
 140 physical storage devices (e.g., disks) should have a copy of each data
 141 block. Default is to use the server-provided default (if any) or 2.
 142 """)
 143
 144 upload_opts.add_argument('--storage-classes', help="""
 145 Specify comma separated list of storage classes to be used when saving data to Keep.
 146 """)
 147
 148 upload_opts.add_argument('--threads', type=int, metavar='N', default=None,
 149                          help="""
 150 Set the number of upload threads to be used. Take into account that
 151 using lots of threads will increase the RAM requirements. Default is
 152 to use 2 threads.
 153 On high latency installations, using a greater number will improve
 154 overall throughput.
 155 """)
 156
 157 run_opts = argparse.ArgumentParser(add_help=False)
 158
 159 run_opts.add_argument('--project-uuid', metavar='UUID', help="""
 160 Store the collection in the specified project, instead of your Home
 161 project.
 162 """)
 163
 164 run_opts.add_argument('--name', help="""
 165 Save the collection with the specified name.
 166 """)
 167
 168 run_opts.add_argument('--exclude', metavar='PATTERN', default=[],
 169                       action='append', help="""
 170 Exclude files and directories whose names match the given glob pattern. When
 171 using a path-like pattern like 'subdir/*.txt', all text files inside 'subdir'
 172 directory, relative to the provided input dirs will be excluded.
 173 When using a filename pattern like '*.txt', any text file will be excluded
 174 no matter where is placed.
 175 For the special case of needing to exclude only files or dirs directly below
 176 the given input directory, you can use a pattern like './exclude_this.gif'.
 177 You can specify multiple patterns by using this argument more than once.
 178 """)
 179
 180 _group = run_opts.add_mutually_exclusive_group()
 181 _group.add_argument('--progress', action='store_true',
 182                     help="""
 183 Display human-readable progress on stderr (bytes and, if possible,
 184 percentage of total data size). This is the default behavior when
 185 stderr is a tty.
 186 """)
 187
 188 _group.add_argument('--no-progress', action='store_true',
 189                     help="""
 190 Do not display human-readable progress on stderr, even if stderr is a
 191 tty.
 192 """)
 193
 194 _group.add_argument('--batch-progress', action='store_true',
 195                     help="""
 196 Display machine-readable progress on stderr (bytes and, if known,
 197 total data size).
 198 """)
 199
 200 run_opts.add_argument('--silent', action='store_true',
 201                       help="""
 202 Do not print any debug messages to console. (Any error messages will
 203 still be displayed.)
 204 """)
 205
 206 _group = run_opts.add_mutually_exclusive_group()
 207 _group.add_argument('--resume', action='store_true', default=True,
 208                     help="""
 209 Continue interrupted uploads from cached state (default).
 210 """)
 211 _group.add_argument('--no-resume', action='store_false', dest='resume',
 212                     help="""
 213 Do not continue interrupted uploads from cached state.
 214 """)
 215
 216 _group = run_opts.add_mutually_exclusive_group()
 217 _group.add_argument('--follow-links', action='store_true', default=True,
 218                     dest='follow_links', help="""
 219 Follow file and directory symlinks (default).
 220 """)
 221 _group.add_argument('--no-follow-links', action='store_false', dest='follow_links',
 222                     help="""
 223 Do not follow file and directory symlinks.
 224 """)
 225
 226 _group = run_opts.add_mutually_exclusive_group()
 227 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
 228                     help="""
 229 Save upload state in a cache file for resuming (default).
 230 """)
 231 _group.add_argument('--no-cache', action='store_false', dest='use_cache',
 232                     help="""
 233 Do not save upload state in a cache file for resuming.
 234 """)
 235
 236 arg_parser = argparse.ArgumentParser(
 237     description='Copy data from the local filesystem to Keep.',
 238     parents=[upload_opts, run_opts, arv_cmd.retry_opt])
 239
 240 def parse_arguments(arguments):
 241     args = arg_parser.parse_args(arguments)
 242
 243     if len(args.paths) == 0:
 244         args.paths = ['-']
 245
 246     args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths]
 247
 248     if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
 249         if args.filename:
 250             arg_parser.error("""
 251     --filename argument cannot be used when storing a directory or
 252     multiple files.
 253     """)
 254
 255     # Turn on --progress by default if stderr is a tty.
 256     if (not (args.batch_progress or args.no_progress or args.silent)
 257         and os.isatty(sys.stderr.fileno())):
 258         args.progress = True
 259
 260     # Turn off --resume (default) if --no-cache is used.
 261     if not args.use_cache:
 262         args.resume = False
 263
 264     if args.paths == ['-']:
 265         if args.update_collection:
 266             arg_parser.error("""
 267     --update-collection cannot be used when reading from stdin.
 268     """)
 269         args.resume = False
 270         args.use_cache = False
 271         if not args.filename:
 272             args.filename = 'stdin'
 273
 274     # Remove possible duplicated patterns
 275     if len(args.exclude) > 0:
 276         args.exclude = list(set(args.exclude))
 277
 278     return args
 279
 280
 281 class PathDoesNotExistError(Exception):
 282     pass
 283
 284
 285 class CollectionUpdateError(Exception):
 286     pass
 287
 288
 289 class ResumeCacheConflict(Exception):
 290     pass
 291
 292
 293 class ResumeCacheInvalidError(Exception):
 294     pass
 295
 296 class ArvPutArgumentConflict(Exception):
 297     pass
 298
 299
 300 class ArvPutUploadIsPending(Exception):
 301     pass
 302
 303
 304 class ArvPutUploadNotPending(Exception):
 305     pass
 306
 307
 308 class FileUploadList(list):
 309     def __init__(self, dry_run=False):
 310         list.__init__(self)
 311         self.dry_run = dry_run
 312
 313     def append(self, other):
 314         if self.dry_run:
 315             raise ArvPutUploadIsPending()
 316         super(FileUploadList, self).append(other)
 317
 318
 319 # Appends the X-Request-Id to the log message when log level is ERROR or DEBUG
 320 class ArvPutLogFormatter(logging.Formatter):
 321     std_fmtr = logging.Formatter(arvados.log_format, arvados.log_date_format)
 322     err_fmtr = None
 323     request_id_informed = False
 324
 325     def __init__(self, request_id):
 326         self.err_fmtr = logging.Formatter(
 327             arvados.log_format+' (X-Request-Id: {})'.format(request_id),
 328             arvados.log_date_format)
 329
 330     def format(self, record):
 331         if (not self.request_id_informed) and (record.levelno in (logging.DEBUG, logging.ERROR)):
 332             self.request_id_informed = True
 333             return self.err_fmtr.format(record)
 334         return self.std_fmtr.format(record)
 335
 336
 337 class ResumeCache(object):
 338     CACHE_DIR = '.cache/arvados/arv-put'
 339
 340     def __init__(self, file_spec):
 341         self.cache_file = open(file_spec, 'a+')
 342         self._lock_file(self.cache_file)
 343         self.filename = self.cache_file.name
 344
 345     @classmethod
 346     def make_path(cls, args):
 347         md5 = hashlib.md5()
 348         md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
 349         realpaths = sorted(os.path.realpath(path) for path in args.paths)
 350         md5.update(b'\0'.join([p.encode() for p in realpaths]))
 351         if any(os.path.isdir(path) for path in realpaths):
 352             md5.update(b'-1')
 353         elif args.filename:
 354             md5.update(args.filename.encode())
 355         return os.path.join(
 356             arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
 357             md5.hexdigest())
 358
 359     def _lock_file(self, fileobj):
 360         try:
 361             fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
 362         except IOError:
 363             raise ResumeCacheConflict("{} locked".format(fileobj.name))
 364
 365     def load(self):
 366         self.cache_file.seek(0)
 367         return json.load(self.cache_file)
 368
 369     def check_cache(self, api_client=None, num_retries=0):
 370         try:
 371             state = self.load()
 372             locator = None
 373             try:
 374                 if "_finished_streams" in state and len(state["_finished_streams"]) > 0:
 375                     locator = state["_finished_streams"][0][1][0]
 376                 elif "_current_stream_locators" in state and len(state["_current_stream_locators"]) > 0:
 377                     locator = state["_current_stream_locators"][0]
 378                 if locator is not None:
 379                     kc = arvados.keep.KeepClient(api_client=api_client)
 380                     kc.head(locator, num_retries=num_retries)
 381             except Exception as e:
 382                 self.restart()
 383         except (ValueError):
 384             pass
 385
 386     def save(self, data):
 387         try:
 388             new_cache_fd, new_cache_name = tempfile.mkstemp(
 389                 dir=os.path.dirname(self.filename))
 390             self._lock_file(new_cache_fd)
 391             new_cache = os.fdopen(new_cache_fd, 'r+')
 392             json.dump(data, new_cache)
 393             os.rename(new_cache_name, self.filename)
 394         except (IOError, OSError, ResumeCacheConflict):
 395             try:
 396                 os.unlink(new_cache_name)
 397             except NameError:  # mkstemp failed.
 398                 pass
 399         else:
 400             self.cache_file.close()
 401             self.cache_file = new_cache
 402
 403     def close(self):
 404         self.cache_file.close()
 405
 406     def destroy(self):
 407         try:
 408             os.unlink(self.filename)
 409         except OSError as error:
 410             if error.errno != errno.ENOENT:  # That's what we wanted anyway.
 411                 raise
 412         self.close()
 413
 414     def restart(self):
 415         self.destroy()
 416         self.__init__(self.filename)
 417
 418
 419 class ArvPutUploadJob(object):
 420     CACHE_DIR = '.cache/arvados/arv-put'
 421     EMPTY_STATE = {
 422         'manifest' : None, # Last saved manifest checkpoint
 423         'files' : {} # Previous run file list: {path : {size, mtime}}
 424     }
 425
 426     def __init__(self, paths, resume=True, use_cache=True, reporter=None,
 427                  name=None, owner_uuid=None, api_client=None,
 428                  ensure_unique_name=False, num_retries=None,
 429                  put_threads=None, replication_desired=None, filename=None,
 430                  update_time=60.0, update_collection=None, storage_classes=None,
 431                  logger=logging.getLogger('arvados.arv_put'), dry_run=False,
 432                  follow_links=True, exclude_paths=[], exclude_names=None):
 433         self.paths = paths
 434         self.resume = resume
 435         self.use_cache = use_cache
 436         self.update = False
 437         self.reporter = reporter
 438         # This will set to 0 before start counting, if no special files are going
 439         # to be read.
 440         self.bytes_expected = None
 441         self.bytes_written = 0
 442         self.bytes_skipped = 0
 443         self.name = name
 444         self.owner_uuid = owner_uuid
 445         self.ensure_unique_name = ensure_unique_name
 446         self.num_retries = num_retries
 447         self.replication_desired = replication_desired
 448         self.put_threads = put_threads
 449         self.filename = filename
 450         self.storage_classes = storage_classes
 451         self._api_client = api_client
 452         self._state_lock = threading.Lock()
 453         self._state = None # Previous run state (file list & manifest)
 454         self._current_files = [] # Current run file list
 455         self._cache_file = None
 456         self._collection_lock = threading.Lock()
 457         self._remote_collection = None # Collection being updated (if asked)
 458         self._local_collection = None # Collection from previous run manifest
 459         self._file_paths = set() # Files to be updated in remote collection
 460         self._stop_checkpointer = threading.Event()
 461         self._checkpointer = threading.Thread(target=self._update_task)
 462         self._checkpointer.daemon = True
 463         self._update_task_time = update_time  # How many seconds wait between update runs
 464         self._files_to_upload = FileUploadList(dry_run=dry_run)
 465         self._upload_started = False
 466         self.logger = logger
 467         self.dry_run = dry_run
 468         self._checkpoint_before_quit = True
 469         self.follow_links = follow_links
 470         self.exclude_paths = exclude_paths
 471         self.exclude_names = exclude_names
 472
 473         if not self.use_cache and self.resume:
 474             raise ArvPutArgumentConflict('resume cannot be True when use_cache is False')
 475
 476         # Check for obvious dry-run responses
 477         if self.dry_run and (not self.use_cache or not self.resume):
 478             raise ArvPutUploadIsPending()
 479
 480         # Load cached data if any and if needed
 481         self._setup_state(update_collection)
 482
 483         # Build the upload file list, excluding requested files and counting the
 484         # bytes expected to be uploaded.
 485         self._build_upload_list()
 486
 487     def _build_upload_list(self):
 488         """
 489         Scan the requested paths to count file sizes, excluding requested files
 490         and dirs and building the upload file list.
 491         """
 492         # If there aren't special files to be read, reset total bytes count to zero
 493         # to start counting.
 494         if not any([p for p in self.paths
 495                     if not (os.path.isfile(p) or os.path.isdir(p))]):
 496             self.bytes_expected = 0
 497
 498         for path in self.paths:
 499             # Test for stdin first, in case some file named '-' exist
 500             if path == '-':
 501                 if self.dry_run:
 502                     raise ArvPutUploadIsPending()
 503                 self._write_stdin(self.filename or 'stdin')
 504             elif not os.path.exists(path):
 505                  raise PathDoesNotExistError("file or directory '{}' does not exist.".format(path))
 506             elif os.path.isdir(path):
 507                 # Use absolute paths on cache index so CWD doesn't interfere
 508                 # with the caching logic.
 509                 orig_path = path
 510                 path = os.path.abspath(path)
 511                 if orig_path[-1:] == os.sep:
 512                     # When passing a directory reference with a trailing slash,
 513                     # its contents should be uploaded directly to the
 514                     # collection's root.
 515                     prefixdir = path
 516                 else:
 517                     # When passing a directory reference with no trailing slash,
 518                     # upload the directory to the collection's root.
 519                     prefixdir = os.path.dirname(path)
 520                 prefixdir += os.sep
 521                 for root, dirs, files in os.walk(path,
 522                                                  followlinks=self.follow_links):
 523                     root_relpath = os.path.relpath(root, path)
 524                     if root_relpath == '.':
 525                         root_relpath = ''
 526                     # Exclude files/dirs by full path matching pattern
 527                     if self.exclude_paths:
 528                         dirs[:] = [d for d in dirs
 529                                    if not any(pathname_match(
 530                                            os.path.join(root_relpath, d), pat)
 531                                               for pat in self.exclude_paths)]
 532                         files = [f for f in files
 533                                  if not any(pathname_match(
 534                                          os.path.join(root_relpath, f), pat)
 535                                             for pat in self.exclude_paths)]
 536                     # Exclude files/dirs by name matching pattern
 537                     if self.exclude_names is not None:
 538                         dirs[:] = [d for d in dirs
 539                                    if not self.exclude_names.match(d)]
 540                         files = [f for f in files
 541                                  if not self.exclude_names.match(f)]
 542                     # Make os.walk()'s dir traversing order deterministic
 543                     dirs.sort()
 544                     files.sort()
 545                     for f in files:
 546                         filepath = os.path.join(root, f)
 547                         # Add its size to the total bytes count (if applicable)
 548                         if self.follow_links or (not os.path.islink(filepath)):
 549                             if self.bytes_expected is not None:
 550                                 self.bytes_expected += os.path.getsize(filepath)
 551                         self._check_file(filepath,
 552                                          os.path.join(root[len(prefixdir):], f))
 553             else:
 554                 filepath = os.path.abspath(path)
 555                 # Add its size to the total bytes count (if applicable)
 556                 if self.follow_links or (not os.path.islink(filepath)):
 557                     if self.bytes_expected is not None:
 558                         self.bytes_expected += os.path.getsize(filepath)
 559                 self._check_file(filepath,
 560                                  self.filename or os.path.basename(path))
 561         # If dry-mode is on, and got up to this point, then we should notify that
 562         # there aren't any file to upload.
 563         if self.dry_run:
 564             raise ArvPutUploadNotPending()
 565         # Remove local_collection's files that don't exist locally anymore, so the
 566         # bytes_written count is correct.
 567         for f in self.collection_file_paths(self._local_collection,
 568                                             path_prefix=""):
 569             if f != 'stdin' and f != self.filename and not f in self._file_paths:
 570                 self._local_collection.remove(f)
 571
 572     def start(self, save_collection):
 573         """
 574         Start supporting thread & file uploading
 575         """
 576         self._checkpointer.start()
 577         try:
 578             # Update bytes_written from current local collection and
 579             # report initial progress.
 580             self._update()
 581             # Actual file upload
 582             self._upload_started = True # Used by the update thread to start checkpointing
 583             self._upload_files()
 584         except (SystemExit, Exception) as e:
 585             self._checkpoint_before_quit = False
 586             # Log stack trace only when Ctrl-C isn't pressed (SIGINT)
 587             # Note: We're expecting SystemExit instead of
 588             # KeyboardInterrupt because we have a custom signal
 589             # handler in place that raises SystemExit with the catched
 590             # signal's code.
 591             if isinstance(e, PathDoesNotExistError):
 592                 # We aren't interested in the traceback for this case
 593                 pass
 594             elif not isinstance(e, SystemExit) or e.code != -2:
 595                 self.logger.warning("Abnormal termination:\n{}".format(
 596                     traceback.format_exc()))
 597             raise
 598         finally:
 599             if not self.dry_run:
 600                 # Stop the thread before doing anything else
 601                 self._stop_checkpointer.set()
 602                 self._checkpointer.join()
 603                 if self._checkpoint_before_quit:
 604                     # Commit all pending blocks & one last _update()
 605                     self._local_collection.manifest_text()
 606                     self._update(final=True)
 607                     if save_collection:
 608                         self.save_collection()
 609             if self.use_cache:
 610                 self._cache_file.close()
 611
 612     def save_collection(self):
 613         if self.update:
 614             # Check if files should be updated on the remote collection.
 615             for fp in self._file_paths:
 616                 remote_file = self._remote_collection.find(fp)
 617                 if not remote_file:
 618                     # File don't exist on remote collection, copy it.
 619                     self._remote_collection.copy(fp, fp, self._local_collection)
 620                 elif remote_file != self._local_collection.find(fp):
 621                     # A different file exist on remote collection, overwrite it.
 622                     self._remote_collection.copy(fp, fp, self._local_collection, overwrite=True)
 623                 else:
 624                     # The file already exist on remote collection, skip it.
 625                     pass
 626             self._remote_collection.save(storage_classes=self.storage_classes,
 627                                          num_retries=self.num_retries)
 628         else:
 629             if self.storage_classes is None:
 630                 self.storage_classes = ['default']
 631             self._local_collection.save_new(
 632                 name=self.name, owner_uuid=self.owner_uuid,
 633                 storage_classes=self.storage_classes,
 634                 ensure_unique_name=self.ensure_unique_name,
 635                 num_retries=self.num_retries)
 636
 637     def destroy_cache(self):
 638         if self.use_cache:
 639             try:
 640                 os.unlink(self._cache_filename)
 641             except OSError as error:
 642                 # That's what we wanted anyway.
 643                 if error.errno != errno.ENOENT:
 644                     raise
 645             self._cache_file.close()
 646
 647     def _collection_size(self, collection):
 648         """
 649         Recursively get the total size of the collection
 650         """
 651         size = 0
 652         for item in listvalues(collection):
 653             if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
 654                 size += self._collection_size(item)
 655             else:
 656                 size += item.size()
 657         return size
 658
 659     def _update_task(self):
 660         """
 661         Periodically called support task. File uploading is
 662         asynchronous so we poll status from the collection.
 663         """
 664         while not self._stop_checkpointer.wait(1 if not self._upload_started else self._update_task_time):
 665             self._update()
 666
 667     def _update(self, final=False):
 668         """
 669         Update cached manifest text and report progress.
 670         """
 671         if self._upload_started:
 672             with self._collection_lock:
 673                 self.bytes_written = self._collection_size(self._local_collection)
 674                 if self.use_cache:
 675                     if final:
 676                         manifest = self._local_collection.manifest_text()
 677                     else:
 678                         # Get the manifest text without comitting pending blocks
 679                         manifest = self._local_collection.manifest_text(strip=False,
 680                                                                         normalize=False,
 681                                                                         only_committed=True)
 682                     # Update cache
 683                     with self._state_lock:
 684                         self._state['manifest'] = manifest
 685             if self.use_cache:
 686                 try:
 687                     self._save_state()
 688                 except Exception as e:
 689                     self.logger.error("Unexpected error trying to save cache file: {}".format(e))
 690         else:
 691             self.bytes_written = self.bytes_skipped
 692         # Call the reporter, if any
 693         self.report_progress()
 694
 695     def report_progress(self):
 696         if self.reporter is not None:
 697             self.reporter(self.bytes_written, self.bytes_expected)
 698
 699     def _write_stdin(self, filename):
 700         output = self._local_collection.open(filename, 'wb')
 701         self._write(sys.stdin, output)
 702         output.close()
 703
 704     def _check_file(self, source, filename):
 705         """
 706         Check if this file needs to be uploaded
 707         """
 708         # Ignore symlinks when requested
 709         if (not self.follow_links) and os.path.islink(source):
 710             return
 711         resume_offset = 0
 712         should_upload = False
 713         new_file_in_cache = False
 714         # Record file path for updating the remote collection before exiting
 715         self._file_paths.add(filename)
 716
 717         with self._state_lock:
 718             # If no previous cached data on this file, store it for an eventual
 719             # repeated run.
 720             if source not in self._state['files']:
 721                 self._state['files'][source] = {
 722                     'mtime': os.path.getmtime(source),
 723                     'size' : os.path.getsize(source)
 724                 }
 725                 new_file_in_cache = True
 726             cached_file_data = self._state['files'][source]
 727
 728         # Check if file was already uploaded (at least partially)
 729         file_in_local_collection = self._local_collection.find(filename)
 730
 731         # If not resuming, upload the full file.
 732         if not self.resume:
 733             should_upload = True
 734         # New file detected from last run, upload it.
 735         elif new_file_in_cache:
 736             should_upload = True
 737         # Local file didn't change from last run.
 738         elif cached_file_data['mtime'] == os.path.getmtime(source) and cached_file_data['size'] == os.path.getsize(source):
 739             if not file_in_local_collection:
 740                 # File not uploaded yet, upload it completely
 741                 should_upload = True
 742             elif file_in_local_collection.permission_expired():
 743                 # Permission token expired, re-upload file. This will change whenever
 744                 # we have a API for refreshing tokens.
 745                 self.logger.warning("Uploaded file '{}' access token expired, will re-upload it from scratch".format(filename))
 746                 should_upload = True
 747                 self._local_collection.remove(filename)
 748             elif cached_file_data['size'] == file_in_local_collection.size():
 749                 # File already there, skip it.
 750                 self.bytes_skipped += cached_file_data['size']
 751             elif cached_file_data['size'] > file_in_local_collection.size():
 752                 # File partially uploaded, resume!
 753                 resume_offset = file_in_local_collection.size()
 754                 self.bytes_skipped += resume_offset
 755                 should_upload = True
 756             else:
 757                 # Inconsistent cache, re-upload the file
 758                 should_upload = True
 759                 self._local_collection.remove(filename)
 760                 self.logger.warning("Uploaded version of file '{}' is bigger than local version, will re-upload it from scratch.".format(source))
 761         # Local file differs from cached data, re-upload it.
 762         else:
 763             if file_in_local_collection:
 764                 self._local_collection.remove(filename)
 765             should_upload = True
 766
 767         if should_upload:
 768             try:
 769                 self._files_to_upload.append((source, resume_offset, filename))
 770             except ArvPutUploadIsPending:
 771                 # This could happen when running on dry-mode, close cache file to
 772                 # avoid locking issues.
 773                 self._cache_file.close()
 774                 raise
 775
 776     def _upload_files(self):
 777         for source, resume_offset, filename in self._files_to_upload:
 778             with open(source, 'rb') as source_fd:
 779                 with self._state_lock:
 780                     self._state['files'][source]['mtime'] = os.path.getmtime(source)
 781                     self._state['files'][source]['size'] = os.path.getsize(source)
 782                 if resume_offset > 0:
 783                     # Start upload where we left off
 784                     output = self._local_collection.open(filename, 'ab')
 785                     source_fd.seek(resume_offset)
 786                 else:
 787                     # Start from scratch
 788                     output = self._local_collection.open(filename, 'wb')
 789                 self._write(source_fd, output)
 790                 output.close(flush=False)
 791
 792     def _write(self, source_fd, output):
 793         while True:
 794             data = source_fd.read(arvados.config.KEEP_BLOCK_SIZE)
 795             if not data:
 796                 break
 797             output.write(data)
 798
 799     def _my_collection(self):
 800         return self._remote_collection if self.update else self._local_collection
 801
 802     def _setup_state(self, update_collection):
 803         """
 804         Create a new cache file or load a previously existing one.
 805         """
 806         # Load an already existing collection for update
 807         if update_collection and re.match(arvados.util.collection_uuid_pattern,
 808                                           update_collection):
 809             try:
 810                 self._remote_collection = arvados.collection.Collection(
 811                     update_collection, api_client=self._api_client)
 812             except arvados.errors.ApiError as error:
 813                 raise CollectionUpdateError("Cannot read collection {} ({})".format(update_collection, error))
 814             else:
 815                 self.update = True
 816         elif update_collection:
 817             # Collection locator provided, but unknown format
 818             raise CollectionUpdateError("Collection locator unknown: '{}'".format(update_collection))
 819
 820         if self.use_cache:
 821             # Set up cache file name from input paths.
 822             md5 = hashlib.md5()
 823             md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
 824             realpaths = sorted(os.path.realpath(path) for path in self.paths)
 825             md5.update(b'\0'.join([p.encode() for p in realpaths]))
 826             if self.filename:
 827                 md5.update(self.filename.encode())
 828             cache_filename = md5.hexdigest()
 829             cache_filepath = os.path.join(
 830                 arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
 831                 cache_filename)
 832             if self.resume and os.path.exists(cache_filepath):
 833                 self.logger.info("Resuming upload from cache file {}".format(cache_filepath))
 834                 self._cache_file = open(cache_filepath, 'a+')
 835             else:
 836                 # --no-resume means start with a empty cache file.
 837                 self.logger.info("Creating new cache file at {}".format(cache_filepath))
 838                 self._cache_file = open(cache_filepath, 'w+')
 839             self._cache_filename = self._cache_file.name
 840             self._lock_file(self._cache_file)
 841             self._cache_file.seek(0)
 842
 843         with self._state_lock:
 844             if self.use_cache:
 845                 try:
 846                     self._state = json.load(self._cache_file)
 847                     if not set(['manifest', 'files']).issubset(set(self._state.keys())):
 848                         # Cache at least partially incomplete, set up new cache
 849                         self._state = copy.deepcopy(self.EMPTY_STATE)
 850                 except ValueError:
 851                     # Cache file empty, set up new cache
 852                     self._state = copy.deepcopy(self.EMPTY_STATE)
 853             else:
 854                 self.logger.info("No cache usage requested for this run.")
 855                 # No cache file, set empty state
 856                 self._state = copy.deepcopy(self.EMPTY_STATE)
 857             if not self._cached_manifest_valid():
 858                 raise ResumeCacheInvalidError()
 859             # Load the previous manifest so we can check if files were modified remotely.
 860             self._local_collection = arvados.collection.Collection(
 861                 self._state['manifest'],
 862                 replication_desired=self.replication_desired,
 863                 put_threads=self.put_threads,
 864                 api_client=self._api_client)
 865
 866     def _cached_manifest_valid(self):
 867         """
 868         Validate the oldest non-expired block signature to check if cached manifest
 869         is usable: checking if the cached manifest was not created with a different
 870         arvados account.
 871         """
 872         if self._state.get('manifest', None) is None:
 873             # No cached manifest yet, all good.
 874             return True
 875         now = datetime.datetime.utcnow()
 876         oldest_exp = None
 877         oldest_loc = None
 878         block_found = False
 879         for m in keep_locator_pattern.finditer(self._state['manifest']):
 880             loc = m.group(0)
 881             try:
 882                 exp = datetime.datetime.utcfromtimestamp(int(loc.split('@')[1], 16))
 883             except IndexError:
 884                 # Locator without signature
 885                 continue
 886             block_found = True
 887             if exp > now and (oldest_exp is None or exp < oldest_exp):
 888                 oldest_exp = exp
 889                 oldest_loc = loc
 890         if not block_found:
 891             # No block signatures found => no invalid block signatures.
 892             return True
 893         if oldest_loc is None:
 894             # Locator signatures found, but all have expired.
 895             # Reset the cache and move on.
 896             self.logger.info('Cache expired, starting from scratch.')
 897             self._state['manifest'] = ''
 898             return True
 899         kc = arvados.KeepClient(api_client=self._api_client,
 900                                 num_retries=self.num_retries)
 901         try:
 902             kc.head(oldest_loc)
 903         except arvados.errors.KeepRequestError:
 904             # Something is wrong, cached manifest is not valid.
 905             return False
 906         return True
 907
 908     def collection_file_paths(self, col, path_prefix='.'):
 909         """Return a list of file paths by recursively go through the entire collection `col`"""
 910         file_paths = []
 911         for name, item in listitems(col):
 912             if isinstance(item, arvados.arvfile.ArvadosFile):
 913                 file_paths.append(os.path.join(path_prefix, name))
 914             elif isinstance(item, arvados.collection.Subcollection):
 915                 new_prefix = os.path.join(path_prefix, name)
 916                 file_paths += self.collection_file_paths(item, path_prefix=new_prefix)
 917         return file_paths
 918
 919     def _lock_file(self, fileobj):
 920         try:
 921             fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
 922         except IOError:
 923             raise ResumeCacheConflict("{} locked".format(fileobj.name))
 924
 925     def _save_state(self):
 926         """
 927         Atomically save current state into cache.
 928         """
 929         with self._state_lock:
 930             # We're not using copy.deepcopy() here because it's a lot slower
 931             # than json.dumps(), and we're already needing JSON format to be
 932             # saved on disk.
 933             state = json.dumps(self._state)
 934         try:
 935             new_cache = tempfile.NamedTemporaryFile(
 936                 mode='w+',
 937                 dir=os.path.dirname(self._cache_filename), delete=False)
 938             self._lock_file(new_cache)
 939             new_cache.write(state)
 940             new_cache.flush()
 941             os.fsync(new_cache)
 942             os.rename(new_cache.name, self._cache_filename)
 943         except (IOError, OSError, ResumeCacheConflict) as error:
 944             self.logger.error("There was a problem while saving the cache file: {}".format(error))
 945             try:
 946                 os.unlink(new_cache_name)
 947             except NameError:  # mkstemp failed.
 948                 pass
 949         else:
 950             self._cache_file.close()
 951             self._cache_file = new_cache
 952
 953     def collection_name(self):
 954         return self._my_collection().api_response()['name'] if self._my_collection().api_response() else None
 955
 956     def manifest_locator(self):
 957         return self._my_collection().manifest_locator()
 958
 959     def portable_data_hash(self):
 960         pdh = self._my_collection().portable_data_hash()
 961         m = self._my_collection().stripped_manifest().encode()
 962         local_pdh = '{}+{}'.format(hashlib.md5(m).hexdigest(), len(m))
 963         if pdh != local_pdh:
 964             self.logger.warning("\n".join([
 965                 "arv-put: API server provided PDH differs from local manifest.",
 966                 "         This should not happen; showing API server version."]))
 967         return pdh
 968
 969     def manifest_text(self, stream_name=".", strip=False, normalize=False):
 970         return self._my_collection().manifest_text(stream_name, strip, normalize)
 971
 972     def _datablocks_on_item(self, item):
 973         """
 974         Return a list of datablock locators, recursively navigating
 975         through subcollections
 976         """
 977         if isinstance(item, arvados.arvfile.ArvadosFile):
 978             if item.size() == 0:
 979                 # Empty file locator
 980                 return ["d41d8cd98f00b204e9800998ecf8427e+0"]
 981             else:
 982                 locators = []
 983                 for segment in item.segments():
 984                     loc = segment.locator
 985                     locators.append(loc)
 986                 return locators
 987         elif isinstance(item, arvados.collection.Collection):
 988             l = [self._datablocks_on_item(x) for x in listvalues(item)]
 989             # Fast list flattener method taken from:
 990             # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
 991             return [loc for sublist in l for loc in sublist]
 992         else:
 993             return None
 994
 995     def data_locators(self):
 996         with self._collection_lock:
 997             # Make sure all datablocks are flushed before getting the locators
 998             self._my_collection().manifest_text()
 999             datablocks = self._datablocks_on_item(self._my_collection())
1000         return datablocks
1001
1002 _machine_format = "{} {}: {{}} written {{}} total\n".format(sys.argv[0],
1003                                                             os.getpid())
1004
1005 # Simulate glob.glob() matching behavior without the need to scan the filesystem
1006 # Note: fnmatch() doesn't work correctly when used with pathnames. For example the
1007 # pattern 'tests/*.py' will match 'tests/run_test.py' and also 'tests/subdir/run_test.py',
1008 # so instead we're using it on every path component.
1009 def pathname_match(pathname, pattern):
1010     name = pathname.split(os.sep)
1011     # Fix patterns like 'some/subdir/' or 'some//subdir'
1012     pat = [x for x in pattern.split(os.sep) if x != '' and x != '.']
1013     if len(name) != len(pat):
1014         return False
1015     for i in range(len(name)):
1016         if not fnmatch.fnmatch(name[i], pat[i]):
1017             return False
1018     return True
1019
1020 def machine_progress(bytes_written, bytes_expected):
1021     return _machine_format.format(
1022         bytes_written, -1 if (bytes_expected is None) else bytes_expected)
1023
1024 def human_progress(bytes_written, bytes_expected):
1025     if bytes_expected:
1026         return "\r{}M / {}M {:.1%} ".format(
1027             bytes_written >> 20, bytes_expected >> 20,
1028             float(bytes_written) / bytes_expected)
1029     else:
1030         return "\r{} ".format(bytes_written)
1031
1032 def progress_writer(progress_func, outfile=sys.stderr):
1033     def write_progress(bytes_written, bytes_expected):
1034         outfile.write(progress_func(bytes_written, bytes_expected))
1035     return write_progress
1036
1037 def desired_project_uuid(api_client, project_uuid, num_retries):
1038     if not project_uuid:
1039         query = api_client.users().current()
1040     elif arvados.util.user_uuid_pattern.match(project_uuid):
1041         query = api_client.users().get(uuid=project_uuid)
1042     elif arvados.util.group_uuid_pattern.match(project_uuid):
1043         query = api_client.groups().get(uuid=project_uuid)
1044     else:
1045         raise ValueError("Not a valid project UUID: {}".format(project_uuid))
1046     return query.execute(num_retries=num_retries)['uuid']
1047
1048 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr,
1049          install_sig_handlers=True):
1050     global api_client
1051
1052     args = parse_arguments(arguments)
1053     logger = logging.getLogger('arvados.arv_put')
1054     if args.silent:
1055         logger.setLevel(logging.WARNING)
1056     else:
1057         logger.setLevel(logging.INFO)
1058     status = 0
1059
1060     request_id = arvados.util.new_request_id()
1061
1062     formatter = ArvPutLogFormatter(request_id)
1063     logging.getLogger('arvados').handlers[0].setFormatter(formatter)
1064
1065     if api_client is None:
1066         api_client = arvados.api('v1', request_id=request_id)
1067
1068     if install_sig_handlers:
1069         arv_cmd.install_signal_handlers()
1070
1071     # Determine the name to use
1072     if args.name:
1073         if args.stream or args.raw:
1074             logger.error("Cannot use --name with --stream or --raw")
1075             sys.exit(1)
1076         elif args.update_collection:
1077             logger.error("Cannot use --name with --update-collection")
1078             sys.exit(1)
1079         collection_name = args.name
1080     else:
1081         collection_name = "Saved at {} by {}@{}".format(
1082             datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
1083             pwd.getpwuid(os.getuid()).pw_name,
1084             socket.gethostname())
1085
1086     if args.project_uuid and (args.stream or args.raw):
1087         logger.error("Cannot use --project-uuid with --stream or --raw")
1088         sys.exit(1)
1089
1090     # Determine the parent project
1091     try:
1092         project_uuid = desired_project_uuid(api_client, args.project_uuid,
1093                                             args.retries)
1094     except (apiclient_errors.Error, ValueError) as error:
1095         logger.error(error)
1096         sys.exit(1)
1097
1098     if args.progress:
1099         reporter = progress_writer(human_progress)
1100     elif args.batch_progress:
1101         reporter = progress_writer(machine_progress)
1102     else:
1103         reporter = None
1104
1105     #  Split storage-classes argument
1106     storage_classes = None
1107     if args.storage_classes:
1108         storage_classes = args.storage_classes.strip().split(',')
1109         if len(storage_classes) > 1:
1110             logger.error("Multiple storage classes are not supported currently.")
1111             sys.exit(1)
1112
1113
1114     # Setup exclude regex from all the --exclude arguments provided
1115     name_patterns = []
1116     exclude_paths = []
1117     exclude_names = None
1118     if len(args.exclude) > 0:
1119         # We're supporting 2 kinds of exclusion patterns:
1120         # 1)   --exclude '*.jpg'    (file/dir name patterns, will only match
1121         #                            the name, wherever the file is on the tree)
1122         # 2.1) --exclude 'foo/bar'  (file/dir path patterns, will match the
1123         #                            entire path, and should be relative to
1124         #                            any input dir argument)
1125         # 2.2) --exclude './*.jpg'  (Special case for excluding files/dirs
1126         #                            placed directly underneath the input dir)
1127         for p in args.exclude:
1128             # Only relative paths patterns allowed
1129             if p.startswith(os.sep):
1130                 logger.error("Cannot use absolute paths with --exclude")
1131                 sys.exit(1)
1132             if os.path.dirname(p):
1133                 # We don't support of path patterns with '..'
1134                 p_parts = p.split(os.sep)
1135                 if '..' in p_parts:
1136                     logger.error(
1137                         "Cannot use path patterns that include or '..'")
1138                     sys.exit(1)
1139                 # Path search pattern
1140                 exclude_paths.append(p)
1141             else:
1142                 # Name-only search pattern
1143                 name_patterns.append(p)
1144         # For name only matching, we can combine all patterns into a single
1145         # regexp, for better performance.
1146         exclude_names = re.compile('|'.join(
1147             [fnmatch.translate(p) for p in name_patterns]
1148         )) if len(name_patterns) > 0 else None
1149         # Show the user the patterns to be used, just in case they weren't
1150         # specified inside quotes and got changed by the shell expansion.
1151         logger.info("Exclude patterns: {}".format(args.exclude))
1152
1153     # If this is used by a human, and there's at least one directory to be
1154     # uploaded, the expected bytes calculation can take a moment.
1155     if args.progress and any([os.path.isdir(f) for f in args.paths]):
1156         logger.info("Calculating upload size, this could take some time...")
1157     try:
1158         writer = ArvPutUploadJob(paths = args.paths,
1159                                  resume = args.resume,
1160                                  use_cache = args.use_cache,
1161                                  filename = args.filename,
1162                                  reporter = reporter,
1163                                  api_client = api_client,
1164                                  num_retries = args.retries,
1165                                  replication_desired = args.replication,
1166                                  put_threads = args.threads,
1167                                  name = collection_name,
1168                                  owner_uuid = project_uuid,
1169                                  ensure_unique_name = True,
1170                                  update_collection = args.update_collection,
1171                                  storage_classes=storage_classes,
1172                                  logger=logger,
1173                                  dry_run=args.dry_run,
1174                                  follow_links=args.follow_links,
1175                                  exclude_paths=exclude_paths,
1176                                  exclude_names=exclude_names)
1177     except ResumeCacheConflict:
1178         logger.error("\n".join([
1179             "arv-put: Another process is already uploading this data.",
1180             "         Use --no-cache if this is really what you want."]))
1181         sys.exit(1)
1182     except ResumeCacheInvalidError:
1183         logger.error("\n".join([
1184             "arv-put: Resume cache contains invalid signature: it may have expired",
1185             "         or been created with another Arvados user's credentials.",
1186             "         Switch user or use one of the following options to restart upload:",
1187             "         --no-resume to start a new resume cache.",
1188             "         --no-cache to disable resume cache."]))
1189         sys.exit(1)
1190     except CollectionUpdateError as error:
1191         logger.error("\n".join([
1192             "arv-put: %s" % str(error)]))
1193         sys.exit(1)
1194     except ArvPutUploadIsPending:
1195         # Dry run check successful, return proper exit code.
1196         sys.exit(2)
1197     except ArvPutUploadNotPending:
1198         # No files pending for upload
1199         sys.exit(0)
1200     except PathDoesNotExistError as error:
1201         logger.error("\n".join([
1202             "arv-put: %s" % str(error)]))
1203         sys.exit(1)
1204
1205     if not args.dry_run and not args.update_collection and args.resume and writer.bytes_written > 0:
1206         logger.warning("\n".join([
1207             "arv-put: Resuming previous upload from last checkpoint.",
1208             "         Use the --no-resume option to start over."]))
1209
1210     if not args.dry_run:
1211         writer.report_progress()
1212     output = None
1213     try:
1214         writer.start(save_collection=not(args.stream or args.raw))
1215     except arvados.errors.ApiError as error:
1216         logger.error("\n".join([
1217             "arv-put: %s" % str(error)]))
1218         sys.exit(1)
1219
1220     if args.progress:  # Print newline to split stderr from stdout for humans.
1221         logger.info("\n")
1222
1223     if args.stream:
1224         if args.normalize:
1225             output = writer.manifest_text(normalize=True)
1226         else:
1227             output = writer.manifest_text()
1228     elif args.raw:
1229         output = ','.join(writer.data_locators())
1230     else:
1231         try:
1232             if args.update_collection:
1233                 logger.info("Collection updated: '{}'".format(writer.collection_name()))
1234             else:
1235                 logger.info("Collection saved as '{}'".format(writer.collection_name()))
1236             if args.portable_data_hash:
1237                 output = writer.portable_data_hash()
1238             else:
1239                 output = writer.manifest_locator()
1240         except apiclient_errors.Error as error:
1241             logger.error(
1242                 "arv-put: Error creating Collection on project: {}.".format(
1243                     error))
1244             status = 1
1245
1246     # Print the locator (uuid) of the new collection.
1247     if output is None:
1248         status = status or 1
1249     elif not args.silent:
1250         stdout.write(output)
1251         if not output.endswith('\n'):
1252             stdout.write('\n')
1253
1254     if install_sig_handlers:
1255         arv_cmd.restore_signal_handlers()
1256
1257     if status != 0:
1258         sys.exit(status)
1259
1260     # Success!
1261     return output
1262
1263
1264 if __name__ == '__main__':
1265     main()